{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,21]],"date-time":"2026-10-21T14:44:17Z","timestamp":1792593857196,"version":"4.5.0"},"reference-count":33,"publisher":"Oxford University Press (OUP)","issue":"1","license":[{"start":{"date-parts":[[2026,1,1]],"date-time":"2026-01-01T00:00:00Z","timestamp":1767225600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/academic.oup.com\/pages\/standard-publication-reuse-rights"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["12071089"],"award-info":[{"award-number":["12071089"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["71991470"],"award-info":[{"award-number":["71991470"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100000265","name":"UK Medical Research Council","doi-asserted-by":"publisher","award":["MRC_MC_UU_00040\/03"],"award-info":[{"award-number":["MRC_MC_UU_00040\/03"]}],"id":[{"id":"10.13039\/501100000265","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,1,20]]},"abstract":"<jats:title>Summary<\/jats:title>\n                  <jats:p>Estimating optimal dynamic treatment regimes (DTRs) from observational data is often complicated by nonignorable missing covariates arising from informative monitoring in clinical practice. A recent weighted Q-learning approach addresses such nonignorable missingness but relies on parametric Q-function models. In multistage settings, misspecification of Q-function models can propagate backwards through pseudo-outcomes and accumulate bias, both directly through Q-function estimation and indirectly through biased estimation of missingness propensities, raising robustness concerns. Motivated by these limitations, we develop robust direct-search estimators of optimal single-stage and multistage treatment rules with nonignorable missing covariates. For single-stage rules, under a future-independent missingness assumption, we construct robust estimators of treatment regime values using covariate balancing weights, augmented with nonparametrically estimated Q-functions. For DTRs, we extend the augmented value estimator by incorporating pseudo-outcomes with inverse missingness propensity weighting, where missingness propensities are modeled semiparametrically using nonresponse instrumental variables. Simulation studies demonstrate that the proposed estimators are more robust to model misspecification and contamination by invalid instrumental variables than weighted Q-learning. An application to electronic medical record data illustrates improved estimation of early fluid resuscitation strategies for sepsis patients when nonignorable missing hemodynamic variables are utilized.<\/jats:p>","DOI":"10.1093\/biostatistics\/kxag041","type":"journal-article","created":{"date-parts":[[2026,10,3]],"date-time":"2026-10-03T11:40:41Z","timestamp":1791027641000},"source":"Crossref","is-referenced-by-count":0,"title":["Robust estimation of optimal dynamic treatment regimes with nonignorable missing covariates"],"prefix":"10.1093","volume":"27","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-7431-7340","authenticated-orcid":false,"given":"Jian","family":"Sun","sequence":"first","affiliation":[{"name":"School of Data Science, Fudan University , Shanghai, 200433,","place":["China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3865-323X","authenticated-orcid":false,"given":"Bo","family":"Fu","sequence":"additional","affiliation":[{"name":"School of Data Science, Fudan University , Shanghai, 200433,","place":["China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0919-3462","authenticated-orcid":false,"given":"Li","family":"Su","sequence":"additional","affiliation":[{"name":"MRC Biostatistics Unit, School of Clinical Medicine, University of Cambridge , Cambridge,CB2 0SR,","place":["United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"286","published-online":{"date-parts":[[2026,10,7]]},"reference":[{"key":"2026102110245457000_kxag041-B1","doi-asserted-by":"publisher","first-page":"673","DOI":"10.1111\/rssb.12129","article-title":"Globally efficient non-parametric inference of average treatment effects by empirical balancing calibration weighting","volume":"78","author":"Chan","year":"2016","journal-title":"J R Stat Soc B"},{"key":"2026102110245457000_kxag041-B2","doi-asserted-by":"publisher","first-page":"3227","DOI":"10.1002\/sim.8659","article-title":"Balancing vs modeling approaches to weighting in practice","volume":"39","author":"Chattopadhyay","year":"2020","journal-title":"Stat Med"},{"key":"2026102110245457000_kxag041-B3","doi-asserted-by":"crossref","DOI":"10.1201\/9781420011180","volume-title":"Missing data in longitudinal studies: strategies for Bayesian modeling and sensitivity analysis","author":"Daniels","year":"2008"},{"key":"2026102110245457000_kxag041-B4","doi-asserted-by":"publisher","first-page":"3503","DOI":"10.1002\/sim.8678","article-title":"Ascertaining properties of weighting in the estimation of optimal treatment regimes under monotone missingness","volume":"39","author":"Dong","year":"2020","journal-title":"Stat Med"},{"key":"2026102110245457000_kxag041-B5","doi-asserted-by":"publisher","first-page":"368","DOI":"10.1080\/01621459.2020.1753522","article-title":"Robust Q-learning","volume":"116","author":"Ertefaie","year":"2021","journal-title":"J Am Stat Assoc"},{"key":"2026102110245457000_kxag041-B6","doi-asserted-by":"publisher","first-page":"1029","DOI":"10.2307\/1912775","article-title":"Large sample properties of generalized method of moments estimators","volume":"50","author":"Hansen","year":"1982","journal-title":"Econometrica"},{"key":"2026102110245457000_kxag041-B7","doi-asserted-by":"publisher","first-page":"353","DOI":"10.1002\/sim.8407","article-title":"Identification of the optimal treatment regimen in the presence of missing covariates","volume":"39","author":"Huang","year":"2020","journal-title":"Stat Med"},{"key":"2026102110245457000_kxag041-B8","doi-asserted-by":"publisher","first-page":"160035","DOI":"10.1038\/sdata.2016.35","article-title":"MIMIC-III, a freely accessible critical care database","volume":"3","author":"Johnson","year":"2016","journal-title":"Sci Data"},{"key":"2026102110245457000_kxag041-B9","doi-asserted-by":"publisher","first-page":"523","DOI":"10.1214\/07-STS227","article-title":"Demystifying double robustness: a comparison of alternative strategies for estimating a population mean from incomplete data","volume":"22","author":"Kang","year":"2007","journal-title":"Stat Sci"},{"key":"2026102110245457000_kxag041-B10","doi-asserted-by":"publisher","first-page":"913","DOI":"10.1093\/jrsssb\/qkad047","article-title":"Non-parametric inference about mean functionals of non-ignorable non-response data without identifying the joint distribution","volume":"85","author":"Li","year":"2023","journal-title":"J R Stat Soc B"},{"key":"2026102110245457000_kxag041-B11","volume-title":"Statistical analysis with missing data","author":"Little","year":"2014","edition":"2nd ed"},{"key":"2026102110245457000_kxag041-B12","doi-asserted-by":"publisher","first-page":"3776","DOI":"10.1002\/sim.7844","article-title":"Augmented outcome-weighted learning for estimating optimal dynamic treatment regimens","volume":"37","author":"Liu","year":"2018","journal-title":"Stat Med"},{"key":"2026102110245457000_kxag041-B13","doi-asserted-by":"publisher","first-page":"692","DOI":"10.1080\/01621459.2018.1537919","article-title":"Estimating dynamic treatment regimes in mobile health using V-learning","volume":"115","author":"Luckett","year":"2020","journal-title":"J Am Stat Assoc"},{"key":"2026102110245457000_kxag041-B14","doi-asserted-by":"publisher","first-page":"3676","DOI":"10.1111\/biom.13872","article-title":"Imputation-based Q-learning for optimizing dynamic treatment regimes with right-censored survival outcome","volume":"79","author":"Lyu","year":"2023","journal-title":"Biometrics"},{"key":"2026102110245457000_kxag041-B15","doi-asserted-by":"publisher","first-page":"331","DOI":"10.1111\/1467-9868.00389","article-title":"Optimal dynamic treatment regimes","volume":"65","author":"Murphy","year":"2003","journal-title":"J R Stat Soc B"},{"key":"2026102110245457000_kxag041-B16","doi-asserted-by":"publisher","first-page":"304","DOI":"10.1007\/s00134-017-4683-6","article-title":"Surviving sepsis campaign: international guidelines for management of sepsis and septic shock: 2016","volume":"43","author":"Rhodes","year":"2017","journal-title":"Intensive Care Med"},{"key":"2026102110245457000_kxag041-B17","doi-asserted-by":"publisher","first-page":"106","DOI":"10.1080\/01621459.1995.10476493","article-title":"Analysis of semiparametric regression models for repeated outcomes in the presence of missing data","volume":"90","author":"Robins","year":"1995","journal-title":"J Am Stat Assoc"},{"key":"2026102110245457000_kxag041-B18","doi-asserted-by":"publisher","first-page":"1096","DOI":"10.1080\/01621459.1999.10473862","article-title":"Adjusting for nonignorable drop-out using semiparametric nonresponse models","volume":"94","author":"Scharfstein","year":"1999","journal-title":"J Am Stat Assoc"},{"key":"2026102110245457000_kxag041-B19","doi-asserted-by":"publisher","first-page":"640","DOI":"10.1214\/13-STS450","article-title":"Q-and A-learning methods for estimating optimal dynamic treatment regimes","volume":"29","author":"Schulte","year":"2014","journal-title":"Stat Sci"},{"key":"2026102110245457000_kxag041-B20","doi-asserted-by":"publisher","first-page":"175","DOI":"10.1093\/biomet\/asv071","article-title":"Semiparametric inverse propensity weighting for nonignorable missing data","volume":"103","author":"Shao","year":"2016","journal-title":"Biometrika"},{"key":"2026102110245457000_kxag041-B21","doi-asserted-by":"publisher","first-page":"4236","DOI":"10.1002\/sim.9857","article-title":"Estimation and evaluation of individualized treatment rules following multiple imputation","volume":"42","author":"Shen","year":"2023","journal-title":"Stat Med"},{"key":"2026102110245457000_kxag041-B22","doi-asserted-by":"publisher","first-page":"4202","DOI":"10.1002\/sim.6223","article-title":"A multiple imputation strategy for sequential multiple assignment randomized trials","volume":"33","author":"Shortreed","year":"2014","journal-title":"Stat Med"},{"key":"2026102110245457000_kxag041-B23","doi-asserted-by":"publisher","first-page":"2338","DOI":"10.1177\/09622802221122397","article-title":"Penalized spline-involved tree-based (PenSIT) learning for estimating an optimal dynamic treatment regime using observational data","volume":"31","author":"Speth","year":"2022","journal-title":"Stat Methods Med Res"},{"key":"2026102110245457000_kxag041-B24","doi-asserted-by":"publisher","first-page":"ujae161","DOI":"10.1093\/biomtc\/ujae161","article-title":"Weighted Q-learning for optimal dynamic treatment regimes with nonignorable missing covariates","volume":"81","author":"Sun","year":"2025","journal-title":"Biometrics"},{"key":"2026102110245457000_kxag041-B25","doi-asserted-by":"publisher","first-page":"723","DOI":"10.5705\/ss.2012.254","article-title":"Empirical likelihood for estimating equations with nonignorably missing data","volume":"24","author":"Tang","year":"2014","journal-title":"Stat Sin"},{"key":"2026102110245457000_kxag041-B26","doi-asserted-by":"publisher","first-page":"2338","DOI":"10.1002\/sim.7643","article-title":"On the use of the not-at-random fully conditional specification (NARFCS) procedure in practice","volume":"37","author":"Tompsett","year":"2018","journal-title":"Stat Med"},{"key":"2026102110245457000_kxag041-B27","doi-asserted-by":"publisher","first-page":"1097","DOI":"10.5705\/ss.2012.074","article-title":"An instrumental variable approach for identification and estimation with nonignorable nonresponse","volume":"24","author":"Wang","year":"2014","journal-title":"Stat Sin"},{"key":"2026102110245457000_kxag041-B28","doi-asserted-by":"publisher","first-page":"199","DOI":"10.1093\/biomet\/asx069","article-title":"Kernel-based covariate functional balancing for observational studies","volume":"105","author":"Wong","year":"2018","journal-title":"Biometrika"},{"key":"2026102110245457000_kxag041-B29","doi-asserted-by":"publisher","first-page":"681","DOI":"10.1093\/biomet\/ast014","article-title":"Robust estimation of optimal dynamic treatment regimes for sequential treatment decisions","volume":"100","author":"Zhang","year":"2013","journal-title":"Biometrika"},{"key":"2026102110245457000_kxag041-B30","doi-asserted-by":"publisher","first-page":"891","DOI":"10.1111\/biom.12836","article-title":"C-learning: a new classification framework to estimate optimal dynamic treatment regimes","volume":"74","author":"Zhang","year":"2018","journal-title":"Biometrics"},{"key":"2026102110245457000_kxag041-B31","doi-asserted-by":"crossref","first-page":"1106","DOI":"10.1080\/01621459.2012.695674","article-title":"Estimating individualized treatment rules using outcome weighted learning","volume":"107","author":"Zhao","year":"2012","journal-title":"J Am Stat Assoc"},{"key":"2026102110245457000_kxag041-B32","first-page":"1","article-title":"Efficient augmentation and relaxation learning for individualized treatment rules using observational data","volume":"20","author":"Zhao","year":"2019","journal-title":"J Mach Learn Res"},{"key":"2026102110245457000_kxag041-B33","doi-asserted-by":"publisher","first-page":"583","DOI":"10.1080\/01621459.2014.937488","article-title":"New statistical learning methods for estimating optimal dynamic treatment regimes","volume":"110","author":"Zhao","year":"2015","journal-title":"J Am Stat Assoc"}],"container-title":["Biostatistics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/biostatistics\/article-pdf\/27\/1\/kxag041\/71668594\/kxag041.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/biostatistics\/article-pdf\/27\/1\/kxag041\/71668594\/kxag041.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,21]],"date-time":"2026-10-21T14:25:01Z","timestamp":1792592701000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/biostatistics\/article\/doi\/10.1093\/biostatistics\/kxag041\/8876144"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1]]},"references-count":33,"journal-issue":{"issue":"1","published-print":{"date-parts":[[2026,1,20]]}},"URL":"https:\/\/doi.org\/10.1093\/biostatistics\/kxag041","relation":{},"ISSN":["1465-4644","1468-4357"],"issn-type":[{"value":"1465-4644","type":"print"},{"value":"1468-4357","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2026,1]]},"published":{"date-parts":[[2026,1]]},"article-number":"kxag041"}}